All this behaviour is suppose to form in most of the time in a natural way...not in a scripted way...
Who said anything about 'scripted'?
I'm talking about game mechanics, and what should or should not be on the table, to make it actually
fun. That's all.
And I'm pointing out that uber-complex behavioral dynamics stuff, while possibly a great idea for a PhD. dissertation or a lifelong attempt to build a really specific kind of AI (lol, the "madness of crowds"), is not really all that fun.
What's fun, about a game where self-directing groups go do various things? Well, let's see, there are a few ways that this could be fun:
1. Players could try to use the same rules and mechanics that the AI does, and grow their own groups, prosper and dominate. Civ game in realtime, basically. Writing the AI would be a major undertaking, but otherwise, fine, there's a game design.
2. Multiple players could try to cooperatively defeat an AI that cheats in various ways, or follows different dynamics. Think "co-op Chicken but strategic scope". Could be really fun and interesting.
3. Multiple players could try and defeat each other and AI agents at the same time. DoTA, basically.
4. Players could play by themselves, cooperatively, or against each other, playing the part of "gods" who can change the gameworld (and thus change the behavior / outcomes of group behaviors). Think, "Populous played with multiple people at once". I think this would be fun, in a frustrating way (no direct control over agents would drive me nuts, personally, no matter how cool and seemingly "smart" they were).
These are just some random ideas.
None of them involve having to simulate Joe the Butcher, who talked to Sally the Gardener, who's sleeping with James the Ruler of the City, who decided that Our Hero(ine) needs to defeat the Wumpus and recover the Thing. That's proper RPG stuff- real storytelling, where a player creates the story to some extent by acting in the way that they think is most appropriate, or is driven along a series of rails, depending on the game design.
We can't really simulate Joe, Sally and James beyond the most rudimentary levels. That's part of what I'm trying to drive home here- there aren't any ways to do that, and I don't think anybody here has a real answer for that atm, even the best work in that area remains a very long way from what you seem to want a "game" to do.
Human behavior is extremely complex, the way it plays out is also complex... and it's better (imo) to abstract it and go for behaviors that don't reflect the whole range of human drama, than to try to create some tapeworm crap where eventually people realize that they're really just poking at a pretty dumb and / or arbitrary set of rules.
Will Wright got a lot of this right when he made The Sims. Maybe go look at that, try to understand what he was doing there, by taking away a lot of the specific stuff but leaving players with the essentials. IIRC, he's on record, talking about how he went about designing a lot of the "interpersonal interactions" in that game, and I suspect that you could probably do so with a lot more actors, if you kept things down to the basic needs of human beings, and then let players poke at the "people" and make things happen. But it wouldn't be a RPG in any real sense- it'd be a simpler Sims with some combat elements, and it would probably not be very interesting. Instead of thinking of everything possible that could be simulated, I really suggest that instead, you think about what would be fun- real fun, not just "hey, this system can do X".
Most games that aren't multiplayer that are fun are doing three things:
1. They're telling you a story. Maybe a multiple-choice story, where you get to make choices that determine the outcome, but it's still a story and you're there for the ride. Plenty of great games in that vein.
2. They're giving you puzzles to solve. This may be anything from physical puzzles (twitch skills) to brain-teasers. Most RPGs mainly puzzle people with practical probability (or how to optimize circumstances to produce a given outcome).
3. They're giving you a memorable experience in some way. This may be visuals, audio, neat character design or aspects of world / actor behavior. It may be stuff like having crowded cities that at least sort've look like the real thing. That's totally doable, so long as the places where it must be faked are hidden away somewhere.
That's pretty much it. Story, challenge, experience. Where the weight is depends on the game. But none of this requires deep-level simulation. 90% of game design past the initial conception is either smoke and mirrors stuff or it's Making Things Work, anyhow. Maybe think more about why you want these self-directing agents, and how players are supposed to be part of a story, be challenged, and have a memorable experience. Maybe players' actions can change the dynamics of individuals, who don't speak, but just emote depending on what the player's doing, as a way to abstract things and keep the tapeworm from being visible? But think through the basics before getting into massive wishlists of detail.
What you're asking for, basically, is not possible. I think that you could simulate a
village of people all acting upon each other and doing things that change the dynamics of the whole village, with the player's actions being key to how things go. That, I am totally certain, is doable. If you keep the behaviors simple, and make the rules totally clear to players, so that they're real participants, not just viewing an ant farm.
But multiple villages, with multiple rulers, which may be building giant empires, based on the behavior of thousands and thousands of agents? I don't think that anything like that exists, beyond a very simplistic level, ala
A-Life projects, etc., and I don't think it would be fun at all, even if you could build one. Just my opinion, of course.